(subtitle: Extending enterprise architecture beyond IT) This presentation (in Notes View, to show slides and script) reviews some of the themes needed to break out enterprise architecture from the usual IT-centric constraints, as represented by most of the existing EA frameworks and toolsets. [Core content copyright (c) Tetradian 2007; other copyrights and trademarks as indicated]

1. Tetradian Consulting
Unit 215, 9 St Johns Street
Colchester CO2 7NN
England
[+44] (0) 781 560 6624
info@tetradian.com
www.tetradian.com
1

2. The current enterprise-architecture frameworks are valuable for reducing IT costs
and complexity. But their over-emphasis on IT, almost to the exclusion of
everything else, can become a hindrance when we need to extend those
architecture principles to the whole of the enterprise.
So the purpose of this slidepack is to identify the limitations of existing
enterprise-architecture frameworks, and to establish criteria for whole-of-
enterprise architecture.
It‟s aimed at people whose work revolves around balancing „big picture‟ issues
with day-to-day practice – roles such as enterprise-architects, process-architects,
strategists and planners – but it should also be useful for senior business
managers.
2

3. As Dana Bredemeyer explains in his influential article, enterprise architecture
grew out of a business need to manage the growing cost and complexity of IT
systems.
As EA maturity developed, it provided new insights into re-purpose and re-use,
and in how to link systems together in new ways – for example, to provide a real-
time view of stock levels on a web-based e-commerce system. More recently the
scope has widened again, with a belated awareness that IT developments must be
linked much more strongly to business needs and business strategy.
The challenge now is to extend this integration to all aspects of the business and
its processes – and even across complex multi-partner enterprises.
3

4. There are quite a few enterprise-architecture frameworks around these days.
Some have a kind of broad „fill in the boxes‟ approach; some work their way
down the technical layers of IT; and others start from more of a process
background, linking across from there to IT. These are some of the better-known
examples of each type.
4

5. The Zachman framework is the „granddaddy‟ of the field, first released almost
twenty years ago. It‟s a straightforward grid, with five layers from most abstract
(„planner‟) to most concrete („sub-contractor‟), and six columns of data, function,
network, people, time and motivation, representing the questions „what?‟,
„how?‟, „where?‟, „who?‟, „when?‟ and „why?‟ respectively.
It‟s close to being the standard reference for the field: every other system maps to
it, and uses it as a check-list for completeness.
5

6. What‟s not so useful is that it‟s only a reference-framework. We can‟t say “We‟re
using the Zachman methodology”, because there isn‟t one: all there is is the grid,
which in practice is just a check-list of possible models. The danger – which we
often see in practice – is that organisations try to „fill in the blanks‟, populating
every possible box in the grid with an appropriate model: it‟s an excellent
academic exercise, of course, but far too expensive in every sense for the realities
of business.
Another problem, which we‟ll also find in all existing frameworks, is that not
only is there an over-emphasis on technical minutiae, but information-technology
is treated as the only source of business knowledge – neglecting the human
derivation of meaning on which organisations ultimately depend. More on this
later.
And like most other frameworks, Zachman places the organisation as the centre
of the world – not in the world. This may seen subtle, but it makes it much harder
to map multi-enterprise processes and exchanges – the kind of complex world
that large organisations now need to manage.
6

7. Dynamic Architecture, or DyA, is another grid-framework, from Dutch
consultancy Sogeti, part of the CAPGemini group.
Like many of the European frameworks, this one is aware of a wider world than
solely IT, but still places most of its emphasis on the technical minutiae – though
oddly only at the abstract level of models. It also covers additional levels above
those of Zachman, namely general principles and policies.
But like Zachman, it‟s still something of a fill-in-the-boxes exercise – though
their methodology does emphasise the need for „just enough, just in time‟
architecture.
7

8. Again unlike Zachman, there is a proper methodology associated with DyA,
although it applies mostly at a rather abstract level.
One very good feature, not seen in most other methodologies, is the systematic
management of „Development without architecture‟ – in other words the projects
and systems that must, for various operational and other reasons, be permitted to
be non-compliant with the architecture.
8

9. The most serious limitation of DyA is that it appears to be proprietary to
CAPGemini. Some parts of the methodology – such as a very useful enterprise-
architecture maturity metric and supporting spreadsheets – are publicly available
from Sogeti‟s website, but the core of the methodology is not.
And despite its brief mention of products and process under the „business
architecture‟ heading, it is, like Zachman, very definitely IT-centric – but doesn‟t
go into much detail even on that.
So a useful „upward and sideways‟ extension of Zachman, but for most purposes
that‟s about it.
9

10. As is perhaps inevitable in a government model, the US Federal Enterprise
Architecture Framework is organised in sets of hierarchies. The framework is
open, detailed – perhaps too detailed in places! – and fully published on the US
government‟s CIO.gov website.
The slide here shows the hierarchy of „reference models‟: performance metrics,
business structures, services, technical structure and, oddly, data-structures
beneath technical.
Note the strange sideways jump between business and services, and again from
services to technical and data...
10

11. ...because the „derivation of value‟ hierarchy shows us what‟s missing, and the
reason for that sideways jump. Information-technology – here shown as
„Technology‟ – is only one of three legs of the model. The other two legs – the
greyed-out boxes for „Human Capital‟ and „Other Fixed Assets‟ – are
placeholders for future work to link these other domains into a true whole-of-
enterprise architecture framework.
11

12. For now, though, FEAF suffers from the same over-emphasis on IT as in the other
frameworks – which is unfortunate, but at least there is an acknowledgement that
there is a world beyond IT.
The other key concern with FEAF is that it‟s so darned big: literally thousands of
pages, much of which may not be relevant outside of a US government context.
And like early iterations of ISO 9000, it‟s all too easy for it to become a
monstrous paper-trail of bureaucracy, consuming vastly more effort than could be
recovered by improved effectiveness.
12

13. TOGAF, The Open Group Architecture Framework, could be described as the
„open source‟ equivalent of FEAF: it‟s not just a government exercise, anyone
can join in with its creation. Currently in its eighth iteration, it uses the same kind
of hierarchy approach as FEAF – though here data-architecture gets a higher
ranking.
As a document, it‟s also a lot more manageable: less than four hundred pages,
more than half of which is the description of the Architectural Design Method, or
ADM methodology.
13

14. The ADM is a well-structured iterative framework, with requirements at the core
– which, among other things, indicates the need for a proper version-controlled
requirements repository as a key part of any architectural process. After each
stage, the repository is updated – which may in turn call for a brief review of a
previous stage, until an overall picture is developed.
The beauty of this approach is that it‟s well suited for what DyA calls „just
enough, just in time‟: a usable skeleton architecture can be developed very
rapidly, with the detail filled in more slowly over time, as needed.
But note that cursory reference to „business architecture‟, near the start of the
cycle; and, by contrast, the enormous bulge of „technical architecture‟...
14

15. ...because TOGAF is not so much IT-centric as IT-obsessed. There‟s no balance at
all: in essence, „business architecture‟ is „anything not-IT‟, all but forgotten in a
headlong rush to drown in the details of technical minutiae – almost exactly what
an enterprise-level architecture aims to avoid, in fact.
And useful though it is, in practice the ADM is little more than a skeleton of
guidelines – it does take a lot of work to flesh it out for any real-world
architecture.
To be fair, the TOGAF community is addressing these issues: for example, the
next release, TOGAF 9, should include much more emphasis on business
integration. But at present, although the framework is excellent for the earlier IT-
centric levels of enterprise-architecture maturity, it‟s actually more of a hindrance
than a help when we need to move beyond those limitations.
15

16. By contrast to their US counterparts, European enterprise-architecture
frameworks have tended to start from product and process, with IT as the
afterthought. ARIS is perhaps the best example: its original product / production
model on the right, and the more recent IT model on the left.
Like TOGAF, requirements are an essential part of each zone of the model, with
processes and services as the centre that holds everything together.
16

17. But strangely, and annoyingly, there doesn‟t seem to be much integration between
process and IT: and the IT base-model is essentially the same as TOGAF, with
much the same limitations around people, „things‟ and narrative knowledge.
Unlike TOGAF, though, the whole thing is „closed‟, proprietary: the models and
methodology are available only through the commercial ARIS toolset – a serious
limitation in itself.
17

18. Finally, to ArchiMate, another European „process-aware‟ framework, like DyA
developed in the Netherlands. „Passive structure‟ is in essence data and its
representation, a kind of expanded equivalent of TOGAF‟s „Data architecture‟;
„active structure‟ shows the equivalent of TOGAF‟s business, applications and
technical architectures; whilst „behaviour‟ is the differing types of process and
service at each level.
Unlike ARIS, it‟s open, in the public domain, already used in several architecture
toolsets such as BizzDesign. And its simple, straightforward structure does map
cleanly to Zachman...
18

19. ...the catch is that‟s about all the structure there is, because ArchiMate is not so
much a methodology as a modelling language, a means to portray architectural
design. In that sense it‟s rather like UML, the Unified Modelling Language so
often used these days in software development.
And its „enterprise architecture‟ portion is, again, derived from the TOGAF
structure, leaving us with the same integration problems as before: everything‟s
IT-centric, with little or no allowance for physical „things‟, for people as people,
or for non-IT-based knowledge.
19

20. But when we have an architecture-framework that does do what we need, we‟ll
then need a plethora of tools and techniques to bring that integration down into
concrete practice. Here at least we are well-served: there are standards and best-
practice frameworks for pretty much everything, from IT governance to
information management, quality management, quality improvement and project
management.
All we‟re missing, then is that high-level framework for enterprise-architecture –
in other words, for a real enterprise-wide architecture.
20

21. As a quick summary, these are the limitations that we face with existing
„enterprise-architecture‟ frameworks.
Without exception, they‟re too IT-centric. Even process-aware frameworks such
as ARIS are poor at understanding people as people, or at modelling flows of
physical „things‟. And none of the frameworks seem to have any concept of the
non-IT-based narrative knowledge that is the primary source of business
meaning.
Even more problematic, the methodologies typically specify just two fixed states
for the architecture: the present, or „as-is‟, and the desirable future, the „to-be‟.
What this convenient partitioning fails to address is that reality is dynamic: by the
time we finally get to the „to-be‟ state, the world has already moved on. We can
perhaps get away with this for a simple technical architecture, but not for a full-
scale enterprise-architecture.
21

22. Equally frustrating is that many of the modelling languages and toolsets reflect
the same limitations.
BPML, the Business Process Modelling Language supported as standard by most
toolsets, has no means to model physical „things: which may be fine for an IT-
automated process, perhaps but not for modelling the equivalent manual process
that‟s required for business continuity when the computer system goes down.
UML is stuck in the same IT-only world: no „things‟, no modelling of people as
people.
And the Business Motivation Model – the supposed standard for motivation
modelling, up in the top-right corner of Zachman – is loaded with similar
limitations, one such being its inability to model a „vision‟ that can be shared
across today‟s complex multi-role, multi-partner enterprises.
The current toolsets for enterprise-architecture, such as System Architect, ARIS
and BizzDesign, end up enforcing these restrictions. Their clumsy user-interfaces
and frequent lack of internal rigour don‟t help in this, either.
22

23. In effect, the current „enterprise-architecture‟ frameworks, languages and toolsets
all portray the world as a kind of „flatland‟, centred on the low-level technical
concerns of IT...
23

24. ...whereas what business needs is a much more rounded view.
Seeing the world from every direction, every dimension.
Not a flatland, but a globe.
And that is what our enterprise-architecture frameworks need to support.
24

25. So here are what I‟d suggest are some core principles for that broader, rounded,
more „global‟ view of the enterprise.
Let‟s look at each of these in a little more detail.
25

26. The existing frameworks tend to bundle everything that‟s not IT into a general
grab-bag labelled „business architecture‟. Before we do anything, we need to
tease these threads apart into something more usable at an enterprise-wide scope.
At an abstract level, it‟s useful to describe the overall enterprise in terms of four
distinct dimensions: business-direction, people, knowledge and physical „things‟.
What we‟d think of as „services‟ and processes sit in the interactions between
these dimensions. Seen this way, it‟s clear that IT is really only one part, one
minor subset, of the overall „knowledge‟ dimension – which helps to bring a
sense of perspective on the limitations of an IT-centric „enterprise architecture‟.
A key point here is that this interaction is dynamic, not static. What makes it
work is a kind of hidden „fifth dimension‟, the systematic integration of the four
dimensions into a coherent whole.
26

27. A tetrahedron provides us with a simple, yet tangible, model of these four
dimensions, the „four corners of the globe‟ from the business perspective.
By rotating our attention constantly between these views, we gain and maintain a
sense of the whole, to keep it working as a whole. This is what enterprise-
architecture really needs to create for us.
(You‟ll find more information about this „tetradian‟ on the Tetradian web-site, at
http://tetradian.com/name .)
27

28. Each area of the business has different emphases on these dimensions – and also
the dimensions they tend to ignore. For example, Operations don‟t have time to
think much about business-directions; IT are perhaps notorious for forgetting
about people, whilst HR perhaps think of little else.
Yet we can‟t expect most business-people to keep all four dimensions in mind at
once, especially under the “Do it now!” pressures of day-to-day business: it‟s way
too much to ask. So it‟s our responsibility, as enterprise architects, to keep that
balance for the enterprise as a whole – maintaining all four dimensions in our
models, always.
28

29. Many business managers are all but obsessed by efficiency; yet efficiency is only
one part of what we really need, which is effectiveness.
There‟s no point in making something more „efficient‟ if it isn‟t reliable. And if it
isn‟t elegant – in both the scientific and human sense – it won‟t work well:
simplicity, clarity and consistency really do matter when maintaining distributed
systems over long periods of time. There‟s also no point in doing something
unless we know it‟s „on purpose‟ – which means we first need to know what the
purpose for that „something‟ is. And we need to ensure that these not only work
together, but tie in well with everything else: without integration, increased
efficiencies in one place invariably lead to greater losses elsewhere in the system.
These may seem obvious: but they can be all too easy to miss unless we test for
them, systematically, repeatedly, in every part of our enterprise-architecture.
(More information about this on the Tetradian website at
http://tetradian.com/strategy#score )
29

30. Existing architecture-frameworks do increasingly call for some kind of linkage to
business drivers and business goals, but don‟t seem to give any guidelines as to
how to do this. In practice, it needs to be one of the core concerns of the „business
direction‟ dimension in the extended architecture.
A key problem is confusion around the term „vision‟. It‟s not a marketing
exercise: true, it needs to be emotive, yet it must also be „larger‟ than the
organisation itself, to create space for customers, partners and so on. A systematic
„motivation audit-trail‟ of vision / role / mission / goal seems to provide the best
approach. (More about this on the Tetradian website at
http://tetradian.com/strategy#vrmg )
Although the vision must be stable, business purpose also needs to be dynamic,
responding to changes in the environment. To track the strategic „weak signals‟ of
future change, a systematic use of foresight theory and practice would seem
important here – likewise a systematic approach to requirements management, of
which more later.
30

31. A key flaw in many existing architecture frameworks is that they assume a final,
unchanging „to-be‟ architecture – like a finished building. But even a building
changes slowly over time, and in any case enterprise-level architecture is more
like the design of a ever-changing city. We need to design for change, not against
it.
Classic frameworks such as FEAF and Zachman tend towards the Waterfall
software model: monolithic, cumbersome, expensive and so slow they‟re out of
date before they start. By contrast, the DyA framework promotes the idea of „just
enough, just in time‟ architecture – the same iterative approach as Agile DSDM
development, which turns conventional software methodology on its head by
making the list of features variable to fit fixed limits of time and budget.
In short, we our architecture to be resilient, self-adaptive, prepared for surprise,
open to experimentation with new technologies – and sufficient support all of this
in a managed way, across the whole enterprise. A hard ask: but it does help to be
clear that it is what we need, rather than holding everyone back at the Waterfall.
31

32. A key concern for enterprise-architecture is the need for governance: there‟s no
point in creating a design if no-one will follow it.
But the „big stick‟ approach rarely works: and architects don‟t have the time or
the clout to police every development in the enterprise. What does work is a
responsibility-based approach: if people „own‟ a project, a data-set, a business-
rule, they‟re more likely to take care of it. A key part of the architect‟s role, then,
is to show people that it‟s in their interest to be responsible on behalf of others as
well as for their own immediate group.
Best tactic: find the local „champions‟ of agile innovation in each area, and make
sure they have the support they need. Help them to keep aligned to the
architecture, and to share ideas and experiences with others, but otherwise keep
out of their way – they‟ll balk if they see anything that seems like control from
the centre!
32

33. As in the TOGAF ADM, requirements are the architecture‟s core. Everything
from the organisation‟s vision to the appearance of a single item on a screen is a
requirement – and all of these are part of the architecture.
The hard part is that they‟re also changing, all the time. (The only requirement
that shouldn‟t change, ever, is the vision.) They‟re dynamic; they belong to
different areas of the enterprise; they conflict. And in an Agile environment,
requirements become more important, not less, because different concerns come
to the fore with each iteration.
Release-management, ownership, version-control, conflict-resolution: these are
all part of requirements-management, and hence for the architecture itself.
33

34. „Service-oriented architecture‟ is something of a buzz-word in IT circles at
present: by describing relationships between systems in a consistent way, as
„services‟, IT complexity can be greatly reduced.
Yet the same applies elsewhere in the enterprise: everything is a „service‟. In
effect, every business-process is a self-contained „system‟ that provides a service;
and each of these services needs clear „service-contracts‟ with its „providers‟ and
„consumers‟, with its own explicit Service Level Agreements, Key Performance
Indicators, Key Success Criteria and the rest.
In this sense, services and service-oriented architecture are the key to creating
simplicity across the whole enterprise – and all of the business benefits that that
would bring.
34

35. It‟s important to understand services first in this abstract way, because it tells us
the nature of each service, and what service-contracts are required with its
„providers‟ and „consumers‟, independent of how the service is implemented.
We then can choose what mix of IT, people-processes and machine-technology to
use in each implementation – in much the same way that each type of soil in a
garden is made up of a different mix of clay, sand and silt.
Each different mix will give the service different SLAs, and different internal
processes; but the overall service – the external service-contracts, the KPIs and
KSCs – should remain the same. Among other advantages, this simplifies
planning for overload and for disaster-recovery: we can change the
implementation of the service - the mix of IT, people and technology – without
changing the service itself.
(More on this on the Tetradian website at http://tetradian.com/entarch#service )
35

36. So let‟s summarise what we‟ve seen in this brief excursion through present
enterprise-architecture.
The current frameworks and tools work well with an IT-centric architecture; but
they don‟t work well for anything else. Dumping everything that‟s not-IT into a
blurry, indistinct grab-bag marked „business-architecture‟ won‟t be acceptable for
much longer.
What we need to do now is expand out that grab-bag into its proper dimensions,
and be clear about their characteristics and the interactions between them,
together with the dynamic nature of business requirements and business
integration. The „four dimensions‟ model gives us the simplest possible
framework for a high-level model of this business „globe‟.
And an expanded concept of abstract „services‟, independent of their
implementations, would provide the key means to apply that high-level model
into purposeful, profitable, business practice.
[ends]
36